Random Forest Classification of 22 GHz Maser Galaxies: A Database Analysis of GBT Observations

To date, approximately 180 extragalactic 22 GHz H2O masers have been detected,
with 115 of these discovered using the Green Bank Telescope (GBT). Of the GBT
detections, 80 were identified by the Megamaser Cosmology Project (MCP), highlight-
ing its significant contribution to the field. These discoveries come from a survey of
approximately 4,600 galaxies using the GBT, putting the rate of discovery at 2.5%.
These rare detections are of exceptional value due to their relevance to precision cos-
mology. They enable direct measurements of supermassive black hole (SMBH) masses
and geometric distances to their host galaxies, both of which are crucial for refining
estimates of the Hubble constant. In order to increase the likelihood of identifying
these valuable galaxies, we developed a machine learning model designed to accurately
classify galaxies based on their detection status. Specifically, we employed a random
forest model that analyzes WISE band colors (W1-W2, W2-W3, W3-W4) and near-
infrared magnitudes (j, h, k) to predict the presence of 22 GHz H2O megamasers. The
accuracy of our model after cross validation averages to approximately 74%.